On Mon, Apr 21, 2008 at 09:26:55PM +0200, Johan Hoffman wrote: > > > > > > Ola Skavhaug wrote: > >> This is general comment on installation of the DOLFIN interface. No > >> doubt, the > >> header files define the interface to DOLFIN. These files are installed > >> somewhere, and we have tools (like pkg-config) to find out how to use > >> the > >> interface (locate, compile, link). Additionally, I will argue that the > >> SWIG > >> interface files also belong to the DOLFIN interface. For a user to > >> extend > >> DOLFIN, and then expose these extensions to Python such that the type > >> information is correct, the SWIG interface files are needed. In SWIG, to > >> propagate type information correctly between separate modules, the > >> %import-directive must be used _on the interface files_. Since these > >> interface > >> files are currently not considered part of the interface, we have no > >> tools > >> (like pkg-config) to resolve the location, or swig-options, of the > >> Python > >> interface. > >> > >> I therefore suggest that we add the swig-interface files when installing > >> DOLFIN. The simplest solution is to install the files the same place as > >> the > >> header files (relatively to the source layout), since the interface > >> files use > >> relative %include. I.e.: > >> > >> $PREFIX/include/dolfin/swig/*.i > >> > >> Comments and objections are welcome. > >> > > > > Sounds reasonable to me as a swig non-expert. > > > > Do we still have the option to -disable-swig/PyDolfin? From previous > discussions it seems that this is desirable since it may complicate the > build process for strict C++ users.
I guess enablePydolfin=no should work. -- Anders _______________________________________________ DOLFIN-dev mailing list [email protected] http://www.fenics.org/mailman/listinfo/dolfin-dev
